Despite high initial start-up costs, service-oriented architecture (SOA) centres of excellence can save organisations a significant amount of money in the longer term and are crucial to ensure service quality, consistency and reuse.
According to analyst Gartner, creating such centres enables enterprises to save an average of 30% in time and expenditure on application integration and data interface development, and cut maintenance outlay by 20%. It can also help them to ensure component reuse levels of about 25%.
But although about 60% of medium-sized and large companies with a turnover of more than £25m have set up some form of centre of excellence, they may not necessarily call it that. Other terms include competency centre, SOA steering committee, central architecture team or project management office.
Less mature organisations may have introduced more informal arrangements instead, creating virtual teams of senior architects and designers who are involved in each SOA project as just one element of their role.
The common theme here is building a group of individuals to become a source of best practice, advice and guidance to ensure the wheel does not have to be reinvented each time a new SOA project is introduced.
An organisation is unlikely to create a centre of excellence to handle its first SOA initiative, but once it moves to its third or fourth, such a notion begins to make sense.
Neil Macehiter, a partner at analyst Macehiter Ward-Dutton, says the issue of when to introduce such a centre tends to be one of scope. "As the scale of the organisation and its ambitions grow, the benefit of a centre is that it can provide centralised guidance in terms of architecture and technical standards, which helps reduce cost and increase value and quality," he says.
So rather than using three different technologies in five different ways across 15 projects, it becomes possible to introduce and enforce a common, consistent, best-practice approach across them all and to manage any deviations from the standard on an exception basis.
This technique not only makes it easier to understand where and why such exceptions occur, it also means personnel do not repeatedly have to go through learning processes that have already been captured elsewhere.
But Ian Finley, a research director at AMR Research, says that although many companies decide to adopt an SOA approach because of the perceived benefits of service re-use, such a situation does not happen spontaneously and has to be planned for.
"To build an architecture rather than just a collection of web services, you need people to sit down and think about what services the business requires, how to build them generically across different situations and how to maintain and deliver them," he says. "It's about discipline and being able to leverage new ideas without people making the same mistake 20 times in 20 different places."
A key challenge is being able to articulate the value in creating such a centre in order to obtain funding for it. The problem is that although centres of excellence require upfront investment, not least in terms of skilled resource, it is notoriously difficult to sell such IT-focused initiatives to the business because they are generally considered to be an overhead. This is particularly true when benefits can seem intangible and a return on investment may not be evident for between 18 months to two years because of the learning curve involved.
As a result, says Paolo Malinverno, a research vice-president at Gartner, many organisations initially undertake a couple of projects to demonstrate that SOA works and, in the process, build up an informal group of architectural and design experts by factoring the costs into any initiatives.
"It's a tough sell, so, typically, people don't just go to a business manager and say they'd like a centre of excellence," he says. "Most are almost introduced by the back door. It's about gradually proving the value and pointing out that this can either be done in a systematic or a piecemeal way, but the latter will be more expensive."
Such a strategy involves introducing metrics to measure before, after and ongoing improvements in service, but can also be quantified in terms of staff requiring less time to get up to speed and/or of reducing initial design costs.
But Macehiter says it can also be valuable to point to other centres of excellence in the organisation, which are generally involved in activities such as customer care or pricing.
"There will be other organisational units that set out to ensure consistency across the business and if people question why they should invest, you can say for the same reason that you invest in a centralised customer service team - to co-ordinate activity and centralise information," he explains.
One organisation that has successfully gone down the SOA centre of excellence route is the Standard Life Group, which provides life assurance, pensions, investment management and healthcare insurance.
The company first set up a so-called "hub-centric" integration centre in 1998 when the decision was made to unify its architecture and provide a standardised framework for the application development team in order to introduce consistency and promote component re-use. The move involved selecting appropriate middleware and selling the vision to the developer community, before introducing 10 infrastructure projects under the new model.
But by 2001 it became clear that to adopt an SOA approach, it would be necessary to make changes at the technology, people and process level to introduce tighter governance and be able to analyse and manage business service impacts more closely.
As part of this process, the centre of excellence, which acts as a shared service, re-factored its framework, which was based on IBM's WebSphere application server, and introduced a Command platform. The platform includes governance and lifecycle management tools, development templates and utilities to facilitate the development process and is updated continually to support new technologies. To date, it has been subject to about 70 person-years of new development and maintenance.
Also, business services, of which there are now about 500, were put into a catalogue or runtime registry to make them easier to find and re-use and a logging mechanism was included to dynamically track which services were being re-employed where and how.
Ian Muir, chief architect at Standard Life, says: "It is a multi-year transition to this architecture and you have to invest in it. These days, you can buy technology off the shelf, but it will still take six to 12 months to get a framework implemented properly and used consistently. After that, it's about building on top of it and investing in how you want to run it. So to get a payback in return-on-investment terms, you're really looking at a two to three-year timeframe."
To encourage consistent usage of the framework requires a multi-pronged approach, however. In Standard Life's case, this involved ensuring that the 10 members of the SOA centre of excellence developed strong relationships with SOA advocates in the development teams, which use Agile methods. Advocates are lead developers who guide their teams in the use of the framework and feed back requirements and ideas to the centre.
But even this is not enough in itself, says Joe Philips, chief architect for the company's SOA centre. "You have to ensure staff build relationships with the right people, but they also have to have the authority to sell the product. It can't be a paper deliverable - it has to be a concrete architecture, design and underlying software framework." But this framework, in turn, must be neither too proscriptive nor too general, he adds.
"You want to promote architectural consistency, but not define how to develop each application for the next 10 years, so there's a skill in pitching it correctly," says Philips. Standard Life, for example, focuses on ensuring "architectural rigour" around the creation of back-end re-usable business services, but gives developers and their business customers control over which front-end user interface they want to employ.
Another important consideration is being able to answer the "what's in it for me?" question to ensure that individual developers are engaged. "We took the approach that a carrot is better than a stick, and so we offer developers tools and automate processes to help them in their day-to-day work," says Philips. "This means they don't have to focus on how things knit together, but can concentrate on building solutions to help the business do what it wants to do."
And the relationship-building activities do not end there, either. The SOA centre team also works alongside a business service management centre of excellence, which is "effectively a virtual police force that ensures quality of service in the catalogue", says Muir. "It's not a full-time team, but is made up of line-of-business people who ensure everyone is working in the same way to the same standards."
Such an approach not only ensures consistency across all the teams, but also means staff can be moved to "where the business demand is, which has the great advantage of creating flexibility of resource", says Muir. Other advantages include business service re-use levels that currently run at about 53%.
But again, it is not enough to simply set up such structures and leave people to get on with it. It is necessary to "constantly sell the benefits to both developers and the business", he adds.
Despite the complexity and effort involved in such activities, not going down this route constantly leads back to the issue of what would work as an alternative, Muir observes.
"Although there is an upfront cost involved, ultimately it has proved cheaper and faster to do it. The question is, where is the crossover point and at what stage does not having a centre of excellence start to hurt you because things aren't flexible enough? That really depends on the complexity of the organisation, but once you get to a certain size, I'd have said it was essential."